A sample report for a made-up mortgage brokerage. The firm and every number in it are invented, to show what you get: each service and each joint measured, what every problem costs you a month, and the order to fix them.
Six services and five joints checked: two services and one joint broken. Most of the rest need attention.
The application form on a phone. About 31 applications a month never reach the CRM. It takes a week to fix, and it is worth more than any other fix in this report.
Each estimate uses your own numbers: 410 leads a month, a 9% close rate and $4,200 of revenue per funded deal.
| Leak | How we worked it out | A month |
|---|---|---|
| Applications from phones that never reach the CRM | 31 a month in GA4 with no matching contact. At your 9% close rate and $4,200 a deal. | $11,700 |
| Phone visitors lost to a slow page and a buried form | If phone visitors applied at three quarters of the desktop rate: 22 more applications a month, at the same close rate. | $8,300 |
| Ad spend on searches that never convert | Spend on search terms with no application in 90 days, from Google Ads. | $2,650 |
| HubSpot seats and contacts you pay for and do not use | Five unused seats, plus the marketing contact tier you would drop. | $450 |
| Lost or at risk, a month | $23,100 | |
Two more problems cost you, but we do not put a number on them: a deal status that is up to a day old, and not being named when people ask an AI assistant. Both are in the report below.
One lead, from a Google search to the lending software, through every joint on its way.
Quick on a desktop, slow on a phone, and the application form is hard to reach on one.
| Measure | Yours | What good looks like | Status |
|---|---|---|---|
| Main content shows on a phone (LCP) | 4.8 s | Under 2.5 s | Broken |
| Response to a tap on a phone (INP) | 380 ms | Under 200 ms | Needs attention |
| Layout shift while loading (CLS) | 0.04 | Under 0.1 | Holding |
| Visitors on a phone | 68% | For context | |
| Application rate on a phone | 1.9% | Close to desktop, 4.3% | Needs attention |
| Application form on the first screen, phone | No | Yes | Needs attention |
The homepage takes 4.8 seconds to show its main content on a typical phone. Two uncompressed images and a chat widget that loads first account for most of it.
Two thirds of your visitors are on a phone, and they apply at less than half the desktop rate. Many leave before the page finishes loading.
Resize and compress the images, load the chat widget after the page, and test again on a mid-range phone over a mobile network.
On a phone the application form starts after two screens of text. The button that scrolls to it is labelled "Learn more".
Visitors ready to apply have to hunt for the form. Every extra scroll loses some of them.
Put a short first step of the form, or a clear "Start your application" button, on the first screen.
Found when people search your name, rarely for what they need, and never cited by AI assistants.
| Measure | Yours | What good looks like | Status |
|---|---|---|---|
| Pages Google has indexed | 64 of 112 | Every page you want found | Needs attention |
| Clicks from search a month | 2,140 | For context | |
| Clicks from people not searching your name | 22% | Over half | Needs attention |
| Target searches in the top 10 | 6 of 40 | Rising every quarter | Needs attention |
| Cited in AI answers | 0 of 25 questions | Cited for your products | Broken |
| Google reviews | 38, rated 4.6 | A request after every funding | Holding |
We asked ChatGPT, Perplexity and Google’s AI Overviews 25 questions your clients ask, such as which broker helps self-employed buyers in Ontario. Three competitors were cited. You were not.
More people ask an assistant before they search. If you are not in the answer, you are not on the shortlist.
Write a page that answers each question plainly, add structured data to product pages, and test the same 25 questions again each quarter.
Your 40 city pages share the same text with the city name swapped. Google treats most of them as duplicates and leaves them out.
The pages meant to bring local searches bring almost none.
Keep the cities you actually serve, give each one real local detail, and merge the rest.
The ad platforms report 212 conversions a month. Your CRM shows 61 applications from ads.
| Measure | Yours | What good looks like | Status |
|---|---|---|---|
| Spend a month, Google and Microsoft | $11,200 | For context | |
| Conversions the platforms count | 212 | Applications, not page views | Broken |
| Applications in the CRM from ads | 61 | For context | |
| Spend on searches that never convert | $2,650 (24%) | Under 10% | Broken |
| Real cost per application | $184 | Tracked every month | Needs attention |
| Ads that send people to the homepage | 60% | Each ad to its own page | Needs attention |
Google Ads counts a conversion whenever the thank-you page loads, including reloads and visits from your own team. It never hears which leads applied or funded.
Bidding chases cheap form fills, not borrowers, and the reports make the campaigns look three times better than they are.
Count applications only, and send applications and funded deals from HubSpot back to Google and Microsoft.
Searches like "mortgage calculator", "mortgage broker jobs" and "what is a HELOC" took $2,650 last month. None led to an application in 90 days.
That money is gone every month, and it crowds out searches from people ready to borrow.
Add negative keywords, split research searches into their own small campaign, and review search terms every week.
You pay for more HubSpot than you use, and its data is hard to trust.
| Measure | Yours | What good looks like | Status |
|---|---|---|---|
| Paid seats used in the last 30 days | 7 of 12 | All | Needs attention |
| Workflows running | 4 | Routing, follow-up and stages at least | Needs attention |
| Contacts with a lead source | 59% | Over 95% | Broken |
| Duplicate contacts | 18% | Under 2% | Broken |
| Open leads with no owner | 312 | None | Needs attention |
| Marketing contacts you pay for | 15,000 tier, 6,900 reachable | Pay for the people you email | Needs attention |
Five paid seats have not logged in for a month. 8,100 of your marketing contacts have unsubscribed, bounced or not opened an email in a year.
About $450 a month for nothing, and every month the contact count creeps toward the next tier.
Remove or reassign the five seats, and set the inactive contacts to non-marketing so you drop a tier.
Four in ten contacts have no lead source, and almost one in five is a duplicate of another record.
You cannot see which channel brings funded deals, so budget moves on guesswork.
Merge duplicates, set lead source automatically from the form and the ad click, and set lifecycle stages by workflow, not by hand.
The CRM and the lending software talk once a night, and sometimes not at all.
| Measure | Yours | What good looks like | Status |
|---|---|---|---|
| How often they sync | Once a night | Within minutes of a change | Needs attention |
| Average delay in a deal’s status | 14 hours | Under 10 minutes | Needs attention |
| Sync runs that failed last quarter | 6 of 92 | None | Broken |
| Failed runs anyone was told about | 0 of 6 | All, the same hour | Broken |
Six nightly runs failed last quarter. Each time, a day of status changes never reached the CRM, and no one was alerted.
Brokers quote a status that is a day or more out of date, and follow-up tasks fire at the wrong time.
Sync on each change rather than once a night, and send an alert to a named person whenever a run fails.
In good order, with one setting to change.
| Measure | Yours | What good looks like | Status |
|---|---|---|---|
| Sender checks (SPF and DKIM) | Pass | Pass | Holding |
| DMARC policy | Monitor only | Quarantine or reject | Needs attention |
| Bounce rate | 1.4% | Under 2% | Holding |
| Replies logged in the CRM | 100% | 100% | Holding |
| Consent on record | Yes | Yes | Holding |
DMARC is set to monitor only, so email that pretends to come from your domain is still delivered.
A fake email in your name can reach clients and lenders, and mailbox providers trust your real email a little less.
Move DMARC to quarantine once the reports show only your own senders, then to reject.
Where one service hands a lead to the next. What we found at each joint, why it costs you, and the fix.
The application form creates a contact in the CRM, but only the name and email arrive. Loan amount, property type and timeline land in one notes field. Leads sent from the form on a phone never arrive at all.
Nobody can sort or route leads by loan size, so a large refinance waits in the same queue as a small question. Every lead sent from a phone is lost before anyone sees it.
Map each form field to its own CRM property and repair the phone form. Then send a test lead through every form, on a desktop and a phone, and check that each field lands on the right record.
The ad platforms count a conversion when someone reaches the thank-you page. They never hear which of those people applied or closed.
The platforms learn to find people who fill in forms, not people who take out loans, so spend drifts toward clicks that never become deals.
Send applications and closed deals from the CRM back to Google Ads and Microsoft Ads, so bidding learns from real outcomes.
Each loan product has its own page, and people who search for a product land on the page for it.
Nothing lost here today: visitors find what they searched for.
No fix needed. Check it again after the next change to the site.
Deal status moves from the lending software into the CRM once a night, in one batch.
All day, the CRM shows yesterday’s status, so brokers quote old information to clients who call to ask where their file is.
Sync status within minutes of each change, and alert the team when a sync fails.
Every reply to an outbound email is logged against the right contact, with consent on record.
Nothing lost here today: nobody calls a prospect twice, and consent is on file.
No fix needed.
Ordered by what each fix is worth for the effort, and by what depends on what.
We read three months of data from your analytics, ad accounts, HubSpot and lending software, with read-only access that you remove when the audit ends. We crawled the whole site, tested every form on a desktop and a phone, and followed real leads from the first click to the lending software.
Money estimates use your own close rate and revenue per deal, and we show the working for each one. Where a problem is real but we cannot measure what it costs, we say so instead of guessing.
This is a sample. The firm is made up, and so is every number and finding. A real report uses your data, your numbers and your systems, and we share it with you privately.
Tell us where your leads come from and where they go next. We reply within one business day, and we say plainly if an audit isn’t the right first step.